Objectives: To be useable in clinical practise, treatments studied in trials must provide sufficient information to enable clinicians and researchers to replicate. We sought to assess the completeness of treatment descriptions in published randomised controlled trials (RCTs) using a checklist and to determine the extent to which peer reviewers and editors comment on the quality of reporting of treatments. Design: A cross-sectional study. Setting: Trials published in the BMJ, a general medical journal. Participants: Fifty-one trials published in the BMJ were independently evaluated by two raters using a checklist. Reviewers' and editors' comments were also assessed for statements on treatment descriptions. Primary and secondary outcome measures: Proportion of trials rated as replicable (primary outcome). Results: For 57% (29/51) of the papers, published treatment descriptions were not considered sufficient to allow replication. Most poorly described aspects were the actual procedures involved including the sequencing of the technique (what happened and when) and the physical or informational materials used (eg, training materials): 53% and 43% not clear, respectively. For a third of treatments, the dose/duration of individual sessions was not clear and for a quarter the schedule (interval, frequency, duration or timing) was not clear. Although the majority of problems were not picked up by reviewers and editors, when they were detected only about two-thirds were fixed before publication. Conclusions: Journals wanting to publish the research of use to practising healthcare professionals need to pay more attention to descriptions of treatments. Our checklist, may be useful for reviewers, and editors and could help ensure that important details of treatments are provided before papers are in the public domain.